J Biol Chem, Vol. 275, Issue 3, 2057-2062, January 21, 2000

Palmitoylation Targets CD39/Endothelial ATP Diphosphohydrolase to Caveolae^*

Katarzyna Koziak, Elzbieta Kaczmarek, Agnes Kittel^§, Jean Sévigny^¶, Jan Krzysztof Blusztajn, Jan Schulte am Esch II^**, Masato Imai, Olaf Guckelberger, Christian Goepfert, Imrana Qawi, and Simon C. Robson

From the Department of Medicine, Center for Immunobiology, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, Massachusetts 02215, the ^§ Hungarian Institute of Sciences, 1450 Budapest, Hungary, and the  Boston University School of Medicine, Boston, Massachusetts 02118

	ABSTRACT

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Ectonucleotidases influence purinergic receptor function by the hydrolysis of extracellular nucleotides. CD39 is an integral membrane protein that is a prototype member of the nucleoside 5'-triphosphate diphosphohydrolase family. The native CD39 protein has two intracytoplasmic and two transmembrane domains. There is a large extracellular domain that undergoes extensive glycosylation and can be post-translationally modified by limited proteolysis. We have identified a potential thioester linkage site for S-acylation within the N-terminal region of CD39 and demonstrate that this region undergoes palmitoylation in a constitutive manner. The covalent lipid modification of this region of the protein appears to be important both in plasma membrane association and in targeting CD39 to caveolae. These specialized plasmalemmal domains are enriched in G protein-coupled receptors and appear to integrate cellular activation events. We suggest that palmitoylation could modulate the function of CD39 in regulating cellular signal transduction pathways.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

The vascular ATP or NTP diphosphohydrolase (ATPDase or NTPDase; EC 3.6.1.5),¹ now known to be CD39, is a plasma membrane-bound enzyme that plays the dominant role in the hydrolysis of extracellular tri- and/or diphosphate nucleotides in blood (1, 2). Our recent data from cd39-null mice indicate that this specific ectonucleotidase also plays a pivotal role in the regulation of an ADP-purinoreceptor P2Y1 function; absence of ATPDase activity results in desensitization of this G protein-coupled receptor with profound effects on hemostasis and thromboregulation (3).

Established topological models of CD39 suggest the presence of two transmembrane domains at both termini of the molecule and an extracellular loop containing a central hydrophobic region (1, 4, 5). The transmembrane domains of ATPDases appear to influence the formation of detergent-sensitive multimers (6). Examination of CD39 amino acid sequences reveals a total of 11 cysteine (Cys) residues, with an unpaired Cys¹³ contained within the intracellular N terminus of the protein (4). Analysis of the CD39 sequence, by a computer algorithm PROSITE, further indicates six putative N-glycosylation sites with several potential casein kinase, cAMP/cGMP-dependent protein kinase or protein kinase C phosphorylation sites (7, 8).

We have described several post-translational modifications of CD39. There are differences in the extent of glycosylation of human CD39 in endothelial cells, platelets, and leukocytes (9). Effects of limited serine proteolysis of native CD39 on ATPDase activity have been documented (5), and there is also a propensity for CD39 to undergo autophosphorylation reactions (data not shown). In addition, we have shown that cellular interactions with free fatty acids modulate ATPDase enzymatic activity in vitro and have postulated that acylation could also influence CD39 structure (10). Here we study this possibility and specifically examine palmitoylation, a reversible and potentially regulated post-translational modification, in which the 16-carbon saturated fatty acid is attached to intracytoplasmic cysteine residues via high energy thioester linkages (11).

Generalized functions for palmitoylation has not been established but many palmitoylated proteins serve as signaling molecules (G protein subunits, -adrenergic receptors, nonreceptor tyrosine kinases, and others) (12, 13). The function of lipid modification of at least some proteins is believed to facilitate membrane association and therefore regulate distribution between plasma membrane and cytoplasm (13-16). There is additional evidence indicating that palmitoylation might direct proteins to caveolae; this may be of relevance as these plasmalemmal microdomains have been implicated in the compartmentalization of signaling molecules (17, 18).

We demonstrate here that CD39 undergoes constitutive palmitoylation within the N-terminal intracytoplasmic region that contains only one potential site for thioester linkage, Cys¹³. This process is associated with strengthened membrane interaction and preferential targeting of the native protein to caveolae. This observation may be pertinent to differential functions, membrane expression and regulation of the increasing number of cell-associated or soluble CD39 family members (8, 19-21).

	EXPERIMENTAL PROCEDURES

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Antibodies and Reagents-- Anti-human CD39 (BU61) monoclonal antibody (mAb) was purchased from AnCell (Bayport, MN). Anti-mouse and anti-rabbit IgG-fluorescein isothiocyanate conjugates, as well as monoclonal mouse M2 anti-FLAG, were purchased from Sigma; peroxidase-conjugated anti-rabbit antibodies were purchased from Pierce. The rabbit polyclonal antibodies used included KY102/130 directed at components of the apyrase conserved region 2 and RO202/217 to apyrase conserved region 4; both have been found to react with human CD39/ATPDase expressed by COS-7 transfectants (5). All biochemical reagents were from Sigma unless otherwise specified and of highest grade available.

Cell Culture-- Human umbilical vein endothelial cells (HUVECs) from fresh umbilical veins were cultured in M199 with 20% fetal calf serum, heparin (100 µg/ml) and endothelial cell growth factor (50 µg/ml) (BioWhittaker, Walkersville, MD), and COS-7-cells were cultured in Dulbecco's modified Eagle's medium with 10% fetal calf serum. Both media were supplemented with L-glutamine (2 mM), penicillin G (100 units/ml), and streptomycin (100 µg/ml. All cells were grown in culture dishes at 37 °C in a humidified incubator with a 5% CO₂ atmosphere. Cultured cells were harvested by scraping; primary cultures of endothelial cells were used at passage 3 (22).

Transient Transfections-- This was performed according to the methods described earlier (1). Briefly, DNA (pcDNA3 vector containing native, N- or C-truncated CD39 cDNA inserts at 85 ng/cm² of the culture plate area) was incubated with LipofectAMINE (Life Technologies, Inc; 3 µg/cm²) for 30 min at room temperature. COS-7 cells, at 80% confluence, were transfected with the DNA/lipofectamine mixture for 5 h at 37 °C in 6 ml of serum-free medium. The transfection was terminated by adding 6 ml of Dulbecco's modified Eagle's medium containing 20% fetal calf serum. Medium was changed after 48 h and cells were harvested for assays after 72 h. Both control COS-7 cells used for transfection and empty vector transfected cells were negative for CD39 as analyzed by Western blotting and flow cytometry with anti-CD39 mAb; in both controls ATPDase activity was negligible.

Stable Transfections-- COS-7 cells were transiently transfected with pcDNA3 vector (Invitrogen, Carlsbad, CA) containing CD39 cDNA as described above. As this vector contains the neomycin resistance gene, stable transfectants were selected by cell culture in medium containing Geneticin (G418; Life Technologies, Inc.) (0.5 mg/ml). Selected clones were grown separately and analyzed for CD39 expression by Western blotting and ATPDase activity assay.

Native and Mutant CD39 Expression Vectors-- For expression of native recombinant CD39 in COS-7 cells, an identical approach to that described previously was utilized (1). Each truncated mutation was generated by polymerase chain reaction using 0.25 µg of a CD39-pcDNA3 vector, described previously (5). Polymerase chain reaction products were gel-cleaned, digested, and ligated into polylinker of the vector utilizing a BamHI site at the 5'-end and a XbaI site at the 3'-end for FLAG-tagged CD39 as well as N- and C-terminal truncations that removed the transmembrane and intracytoplasmic domains at either site. The pcDNA3 vectors were utilized for the transmembrane domain truncations and the FLAG-tagged native protein. To assure attachment of a FLAG tag in C-terminal truncated forms and FLAG-tagged CD39, the antisense primer used for polymerase chain reaction included the nucleotide sequence coding for FLAG. Constructs in expression vectors were sequenced using an Applied Biosystems 373 Fluorescent DNA sequencer (Perkin-Elmer, Foster City, CA) using SP6 and T7 primers as well as custom sequencing primers (5).

Cell Labeling and Immunoprecipitation-- Forty-eight h post-transfection, COS-7 cells were washed with serum-free Dulbecco's modified Eagle's medium and then incubated with 0.4 mCi/ml of [9,10-³H]palmitic acid for 1 h at 37 °C. Cells were washed twice in ice-cold Tris-saline buffer (50 mM NaCl, 20 mM Tris, pH 8.0), harvested by scraping, and then lyzed in buffer containing 20 mM Tris, pH 7.5, 150 mM NaCl, 1 mM phenylmethylsulfonyl fluoride, 0.02 KIU/ml aprotinin, 4 µg/ml leupeptin, 2 mM sodium orthovanadate, 10 mM sodium pyrophosphate, 0.1 M sodium fluoride, and 1% Nonidet P40 (Nonidet P-40). Cell lysates were clarified by centrifugation at 10,000 × g for 10 min at 4 °C. The supernatants were precleared by 1-h incubation with 10 µl of protein A/G-agarose at 4 °C. Labeled CD39 was immunoprecipitated with 3 µg of anti-CD39 mAb (overnight at 4 °C or with equivalent levels of anti-FLAG mAb; M2, Eastman Kodak Co.), followed by 2-h incubation with 50 µl of protein A/G-agarose at 4 °C. Immunoprecipitates were washed five times with lysis buffer, eluted with 2 × nonreducing Laemmli sample buffer by boiling for 5 min and analyzed by SDS-PAGE (23).

Western Blotting and Autoradiography-- Immunoprecipitates or conditioned media (100 µl, prelyophilization) were separated on 10% SDS-PAGE, including a 4% stacking gel or 4-15% gradient gels (Bio-Rad) and transferred to polyvinylidene difluoride membrane (Immobilon-P, Millipore, Bedford, MA) by electroblotting (24). Proteins were probed with mAbs to human CD39 (BU61) and to the FLAG epitope or with polyclonal rabbit antibodies raised against the CD39-derived peptides amino acids 102-130 (KY102/130) or amino acids 202-217 (RO202/217) (5). Protein bands were visualized using horseradish peroxidase-conjugated goat anti-rabbit IgG or goat anti-mouse IgG (Pierce) followed by the Renaissance Enhanced Chemiluminesence reagent plus (ECL, NEN Life Science Products) according to the manufacturer's instructions. The gel for autoradiography was fixed for 30 min in 30% methanol, 10% glacial acetic acid solution, soaked for 30 min in Enhancer (NEN Life Science Products), dried, and exposed to Kodak film for 21 days at 80 °C.

ATPDase Assays-- Enzyme activity was determined at 37 °C in 1 ml of 8 mM CaCl₂, 200 µM substrate (ATP or ADP), and 50 mM Tris imidazole, pH 7.5. Reactions were terminated with 0.25 ml of malachite green reagent. Released [P_i] was determined and validated as described previously (25).

Thin Layer Chromatography (TLC)-- As radiolabeled palmitic acid may be metabolized to myristate and confound the experimental conclusion, we confirmed the nature of the acylation of CD39 as follows. Cell lysates from transient transfections with native CD39 were immunoprecipitated with anti-human CD39, treated with either 1 M hydroxylamine, pH 7.5, to remove the incorporated radiolabel from CD39 (and from whole lysate); in parallel, control immunoprecipitates were also incubated with 1 M Tris-HCl, pH 7.5, as described previously (15). Both control and the alkaline hydrolysis reactions were carried out overnight at room temperature. Released fatty acids were extracted into chloroform:methanol:water (2:1:1), dried under nitrogen, and then analyzed by C-18 reverse phase TLC (Whatman Inc.). The positions of standards [³H]myristic acid and [³H]palmitic acid or the test samples were observed after development in acetic acid:acetonitrile (1:1). The amount of radioactivity was determined using a Bioscan 2000 radiochromatogram scanner.

Enzyme Cytochemistry-- For the demonstration of ATPDase activity, a cerium precipitation method was used (22). After discarding the medium and washing with cacodylate buffer (0.25 M sucrose in 0.05 M sodium cacodylate, pH 7.4), transfected cells and primary cultures of HUVECs were fixed with ice-cold cacodylate buffer containing 3% paraformaldehyde (Merck, Darstadt, Germany), 0.5% glutaraldehyde (Taab, Aldermaston, Berks, United Kingdom) and 2 mM CaCl₂, pH 7.4, for 30 min. After several rinses in 0.07 M Tris-maleate buffer, pH 7.4, the cells were incubated in a medium containing either ATP or ADP (1 mM) as the relevant substrate, 2 mM CeCl₃ (capture agent for the liberated P_i), 1 mM levamisole (inhibitor of alkaline phosphatases; Amersham Pharmacia Biotech, Buckinghamshire, UK), 1 mM ouabain (Na⁺/K⁺-ATPase inhibitor; Merck, Darmstadt, Germany), 50 µM ,-methylene-ADP (5'-nucleotidase inhibitor) and KCl (5 mM), in Tris-maleate buffer (70 mM, pH 7.4) for 30 min at 37 °C. This incubation was followed by three rinses in Tris-maleate buffer, and samples were then post-fixed for 30 min in 1% O_sO₄ (Taab, Aldermaston, Berkshire, UK) dissolved in cacodylate buffer, dehydrated, and embedded in Epon (Fluka, Buchs, Switzerland). Ultra-thin sections were cut and examined in a Hitachi 2001 transmission electron microscope (Hitachi Corp., Tokyo, Japan). To demonstrate the specificity of the reaction product, in control experiments, the relevant substrate (ATP or ADP) was omitted from the incubation medium.

Immunostaining with Anti-CD39-- Fixation was carried out as above and after rinsing with PBS, HUVECs or transfected cells were blocked with 5% control human serum in PBS for 30 min and incubated overnight with the antibodies (CD39 mAb in 1:500 dilution) on a shaking plate at 4 °C. The Vectastain ABC system was used according to the manufacturer's instruction (Vector Lab, Burlinghame, CA). Post-fixation, embedding, and electron microscopic investigations were performed as for enzymatic histochemical staining.

Membrane Association of CD39 and Mutants-- Aliquots of conditioned supernatant fluids were collected from all transfected COS-7 cells, ultracentrifuged for 60 min at 100,000 × g and 4 °C. Supernatants were then freeze-dried and concentrated specimens dissolved by boiling in 2 × nonreducing Laemmli sample buffers. Samples (equivalents of 100 µl of conditioned medium) were analyzed by Western blotting with mAb to CD39, and bands were visualized with ECL as described above.

	RESULTS

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

Palmitoylation of CD39 and Mutant Forms-- Following incubation of transfected COS-7 cells with [³H]palmitic acid, anti-CD39 (or anti-FLAG; not shown) mAb was used to immunoprecipitate radiolabeled proteins from cell lysates. The subsequent analysis of the immunoprecipitated protein by Western blotting with BU61 and other anti-CD39 established that the transfected COS-7 cells expressed the tagged CD39, N- (1-37) and C-truncated (477-510) proteins at near equivalent levels (Fig. 1A).

View larger version (81K): [in this window] [in a new window]   Fig. 1.   A, Western blot analysis of COS-7 cells transiently transfected with pcDNA3-CD39 and N- and C-truncated CD39 mutants. Proteins were immunoprecipitated with anti-CD39 (or anti-FLAG; not shown), separated on 4-15% SDS-PAGE, immunostained with anti-human CD39 Ab, and visualized by ECL. B, autoradiography of COS-7 cells transiently transfected with CD39 and N- and C-truncated mutants incubated with [13C]palmitic acid. Cell lysates were immunoprecipitated with anti-human CD39 Ab (BU61) followed by SDS-PAGE and autoradiography (see "Experimental Procedures").

In parallel, we also observed, by autoradiography, radioactive signals indicating incorporation of the [³H]palmitic acid label into native CD39 and the C-truncated mutant, but not the N-truncated protein (Fig. 1B). There were no radiolabeled bands detected for empty vector transfected cells (data not shown). Palmitoylation of CD39 expressed in HUVECs was also demonstrated (data not shown). These results suggested that the Cys¹³ present within the intracytoplasmic domain at the N terminus of CD39 underwent palmitoylation.

However, palmitic acid can be metabolized into myristate by cells in culture (15). Therefore, to confirm the nature of the protein acylation, CD39-transfected cell lysates were treated with either 1 M hydroxylamine, pH 7.5, to cleave thioester bonds linking fatty acid (palmitic acid or its putative metabolite, myristic acid) to the protein backbone or with 1 M Tris-HCl, pH 7.5, as the parallel control. Hydrolyzed fatty acids were then extracted into chloroform:methanol:water (2:1:1), dried under nitrogen, and analyzed by C-18 reverse phase TLC (Fig. 2). The positions of standards ([³H]myristic acid (MA) and [³H]palmitic acids (PA)) and test samples were determined after development in acetic acid:acetonitrile (1:1). Both standard fatty acids and other labeled membrane lipids were detected and readily separated by TLC; evidence of conversion of the labeled palmitic acid to myristate in vitro was demonstrated by analysis of the total transfected cell lysate fractions hydrolyzed with either hydroxylamine or incubated with Tris-HCl (Fig. 2, A and B, respectively).

View larger version (35K): [in this window] [in a new window]   Fig. 2.   Thin layer chromatography separation. Fatty acids released from either whole cell lysates of COS-7 cells transiently transfected with native CD39 (A, B) or cell lysates immunoprecipitated with anti-CD39 (C, D) and hydrolyzed with either 1 M hydroxylamine, pH 7.5 (A, C), or 1 M Tris-HCl, pH 7.5, as a control (B, D). PA, palmitic acid; MA, myristic acid.

We further evaluated the specific nature of modification of CD39 by performing first the immunoprecipitation with anti-CD39 followed by incubation of the entire product with either 1 M hydroxylamine or 1 M Tris-HCl, as described above. Examination of radiolabeled released fatty acids by TLC clearly identified that palmitic acid was incorporated into CD39, confirming that the ectonucleotidase was only subject to palmitoylation. This acylation appeared to be via thioester linkages to the Cys¹³ within the N-terminal domain (Fig. 2C). In the parallel control experiment, no palmitic acid was released (Fig. 2D).

Membrane Association of CD39 and Truncated Mutants-- We then investigated the membrane association of CD39 and mutants to evaluate potential release of the truncated proteins from the cell membrane. Conditioned media free of cellular debris were collected from COS-7 cultures transfected with native, N and C terminus truncated mutants of CD39. The deletion of the N-terminal region of CD39 and consequent associated lack of acylation resulted in some release of this mutant to the medium, unlike the native protein or C-terminal truncated mutant that were strictly cell-associated (Fig. 3). We estimated that less than 10% of the immunoreactive N-terminal mutant antigen was solubilized (not shown).

View larger version (14K): [in this window] [in a new window]   Fig. 3.   Western blot analysis of conditioned medium. COS-7 cells were transiently transfected with pcDNA3-CD39, N- or C-truncated CD39 mutants. Only the N-terminal truncated mutant protein was detected in a soluble state in the concentrated conditioned medium of transfected cells.

However, this observation suggested that the N terminus might play some regulatory role in the attachment of CD39 to the cell membrane. We further examined the effects of mutagenesis and deletion of the intracytoplasmic domains on CD39 membrane distribution patterns by additional immunohistochemistry and cytochemical studies.

Electron Microscopy and Cytochemistry-- Functional CD39 expression was determined by visualization of CD39 and mutant ectonucleotidase ATPase activity in transiently transfected COS-7 cells. Empty vector-transfected cells had no specific ATPase activity (Fig. 4A). The pattern of membrane expression of native recombinant CD39 was remarkable in that ATPase activity was largely concentrated in plasmalemmal microdomains, highly suggestive of caveolar structures (Fig. 4B), and in keeping with our prior morphological observations (26). The C-terminal truncated mutant also appeared to be enriched in the caveolae, whereas the N-terminal mutant did not localize in these domains and was only observed in decreased amounts within the plasma membrane (Fig. 4, C and D, respectively).

View larger version (142K): [in this window] [in a new window]   Fig. 4.   Electron microscopy of COS-7 cells transiently transfected with pcDNA3 vector (A), pcDNA3-CD39 (B), C-truncated (C), and N-truncated CD39 mutants (D). Magnification: × 45,000. Small black dot-like structures indicated by arrows are caveolae and stain for ATPase activity readily visualized in only the CD39 and C-terminal mutant transfected cells. The "asterisk" symbol designates the extracellular space.

HUVECs were also shown to express high levels of immunoreactive and enzymatically active CD39 in caveolae-like structures (Fig. 5, A and B). We have previously demonstrated that antibodies to CD39 and caveolin both react with these vesicles, confirming specific localization of CD39 within caveolae (26). A similar distribution of CD39-antigen and associated specific ATPase activity was shown in stable transfectants expressing CD39 (Fig. 5, C and D, respectively).

View larger version (138K): [in this window] [in a new window]   Fig. 5.   Electron microscopy of HUVECs (A, B) and COS-7 cells stable transfectants with native pcDNA3-CD39 (C, D) stained with either anti-CD39 mAb (A, C) or for ATPase activity (B, D). Magnifications: × 8,000 (A), × 2,600 (B), 10,000 (C), and 20,000 (D). As before, the small black dots, some of which are indicated by arrows, correspond to caveolae, and the "asterisk" indicates the extracellular space.

	DISCUSSION

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES

In this manuscript, we demonstrate that CD39 could undergo a process of S-acylation and show that labeled free palmitic acid could associate with the native protein by a thioester bond, as observed for other membrane proteins (12, 13, 15). Truncated forms of CD39 lacking the N-terminal intracytoplasmic region, and the associated Cys¹³ residue (5), were not subject to palmitoylation. This process of palmitoylation in the native protein appeared to be constitutive and to contribute to the integral membrane association of this ectonucleotidase. Thus, a small proportion of this N-terminal truncated CD39 protein was also found in conditioned medium of transfected cells in a soluble form (Fig. 3). The apparent targeting of CD39 to caveolae also appeared linked to this post-translational modification as the mutant lacking this potential palmitoylation site was not detected in these plasmalemmal microdomains (Fig. 4D).

Because the hydrolysis of palmitic acid from CD39 could occur in an alkaline environment, it was presumed that the linkage was a thioester bond (15, 27). We found no evidence for myristoylation of CD39 (Fig. 2C), and there are no putative N-terminal glycine residues that would be accessible for amide N-acylation linkages (4). Hence, unlike the majority of acylated membrane proteins, palmitic acid substitution of CD39 did not require prior modification by myristoyl groups. Nor was there any potential for prenylation at the C-terminal region, given the absence of strict consensus sequences in this region (13). Other proteins that are only palmitoylated are generally targeted to the inner surface of the plasma membrane and include G_q family members (12, 14) and neuronal proteins, e.g. neuromodulin and SNAP-25 (11).

In general, the functional significance of palmitoylation remains unclear. In part, this relates to technical limitations with respect to the generation of mutants, that preclude palmitoylation by conversion of Cys residues or by deletion mutagenesis. These changes may also influence protein structure ab initio (13). However, alternative approaches are limited in that proposed inhibitors of palmitoylation, such as tunicamycin (28), would be concomitant inhibitors of CD39 glycosylation (5). Given this caveat, it remains feasible that palmitoylation could influence the association of CD39 with the cell membrane (29) and could also facilitate interactions with other acylated proteins, such as caveolin (30).

The potential multimerization of CD39 (6) may be facilitated by acylation with intermolecular interactions within the cholesterol and sphingolipid-rich caveolar microdomains of the plasma membrane (30). The formation of multimers of membrane proteins has been also observed for the seven-transmembrane domain G protein-coupled receptors; these undergo C-terminal palmitoylation (13). Proteins with N-terminal acylation, related to the GTPases, can be also potentially linked to purinergic signaling events (31, 32). Such intrinsic protein structural changes could all be linked to the promotion of functional multimeric units that modulate biochemical activity. In this regard, we have previously shown that supplementation of endothelial cell cultures with exogenous palmitic acid may boost ATPDase activity by approximately 50%; however, whether this is a direct effect on the CD39 or operative through modulation of membrane lipids remains unclear (10, 33). Moreover, membrane cholesterol is essential for caveolar function (34). Thus, pathological changes in cholesterol homeostasis have the potential to adversely influence vascular ATPDase activity, perturb other thromboregulatory proteins localized within caveolae, e.g. eNOS (16, 35, 36), with implications for atherosclerosis and platelet sequestration within abnormal vessels (37).

Biologically active CD39 is expressed mainly in the plasma membrane and is subject or responsive to different forms of surface stimulation (10, 22). Although we have not yet examined this in detail, there is the possibility that CD39 may be recycled to and from cell membranes via sequential actions of palmitoyltransferases and palmitoyl-protein thioesterases (13).

The biological properties of caveolae appear to bolster these speculations. These plasmalemmal microdomains have the potential to compartmentalize signaling responses and are enriched in proteins intimately associated with the regulation of calcium flux and signaling pathways (38). Our data imply that both CD39 antigen and the associated ATPase activity are associated with caveolae (Figs. 4 and 5). The targeting of palmitoylated CD39 to caveolae could influence defined G protein-coupled receptors within this plasmalemmal microenvironment (17).

We show that the prototype CD39 family member undergoes palmitoylation at the N-terminal intracytoplasmic domain and is also targeted to caveolae. This post-translational modification would not be anticipated in CD39L1 but may also occur in other CD39-like proteins, e.g. CD39L3 has putative palmitoylation sites within both the N- and C-intracytoplasmic regions (20, 39). Other members of the CD39 family have only a single transmembrane sequence at the N terminus that, in the case of the macrophage-associated CD39L4, undergoes cleavage to result in a soluble enzyme (21). How any of these modifications influence the pathophysiological function(s) of the CD39/E-NTPDase family members remains to be determined.

	ACKNOWLEDGEMENTS

We thank Dr. Susan Hagen for the access to electron microscopy equipment at the Center for Advanced Microscopy, Beth Israel Deaconess Hospital and for her generous help and advice. We thank Eva Csizmadia for endothelial cell culture and Piotr Kaczmarek for preparation of illustrations.

	FOOTNOTES

^* This work was supported in part by National Institutes of Health Grants R01 HL57307 (to S. C. R.) and P01 AG09525 (to J. K. B.) and American Heart Association Grant-in-aid 9650490N (to S. C. R.). A poster demonstrating the caveolar localization of CD39 was presented by Agnes Kittel et al at the 2nd International Workshop on Ecto-ATPases and Ectonucleotidases, Diepenbeek, Belgium in June 1999.The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Joint first authors.

^¶ Recipient of studentships from the Heart and Stroke Foundation of Canada and "Fonds pour la Formation de Chercheurs et 1'Aide à la Recherche du Québec."

^** Funded by a postdoctoral fellowship from the Deutsche Forschungsgemeinschaft.

To whom correspondence should be addressed: Rm. 370 H, Research North, Beth Israel Deaconess Medical Center, 99 Brookline Ave., Boston, MA 02215. Tel.: 617-632-0831; Fax: 617-632-0880; E-mail: srobson@caregroup.harvard.edu.

	ABBREVIATIONS

The abbreviations used are: ACR, apyrase conserved regions; ATPDase (or NTPDase), ATP (or NTP) diphosphohydrolase (now known to be CD39); EC, endothelial cell; mAb, monoclonal antibody.

	REFERENCES

TOP ABSTRACT INTRODUCTION EXPERIMENTAL PROCEDURES RESULTS DISCUSSION REFERENCES